The NRS-RT scale: A dedicated nutritional risk screening tool for patients undergoing radiotherapy - Initial...

Cancer-associated malnutrition is a major determinant of impaired treatment tolerance and adverse clinical outcomes in patients receiving radiotherapy (RT). Despite this, no RT-specific nutritional screening instrument has been established. Herein, we report the initial single-center development and validation of a dedicated Nutritional Risk Screening Tool for Radiotherapy (NRS-RT).

A two-phase sequential study design was adopted. Phase 1 (January 2024 to August 2024) comprised two rounds of Delphi consensus consultations (20 experts in Round 1, 13 in Round 2) to develop the scale. Phase 2 (September 2024 to December 2025) prospectively recruited 254 histopathologically confirmed cancer patients scheduled for RT. The screening and assessment performance of NRS-RT was compared against the reference standards NRS-2002 and PG-SGA. Confirmatory factor analysis (CFA), bootstrap 95% confidence intervals (1,000 resamples), and sensitivity analyses were also performed.

The final NRS-RT consists of 19 items (0-100 points) and exhibits acceptable internal consistency (Cronbach's α = 0.778). Exploratory factor analysis initially extracted six factors with eigenvalues > 1.0; the sixth factor showed substantial cross-loadings and was not clinically stable, so a five-factor structure explaining 60.74% of the total variance was retained. NRS-RT demonstrated excellent agreement with NRS-2002 (κ = 0.904 (95% CI: 0.861-0.947), sensitivity 92.9% (95% CI: 87.1-96.2%), specificity 97.2% (95% CI: 93.0-99.0%)) and good agreement with PG-SGA (weighted κ = 0.823 (95% CI: 0.783-0.878)), with an AUC of 0.990 at the first-stage ROC against PG-SGA (Stage 1) and an optimal cutoff value of 7 points for detecting nutritional risk. Discordant patients missed by NRS-2002 or underestimated by PG-SGA had high rates of clinically significant weight loss (>5%) during RT (42.9% and 41.7%, respectively).

NRS-RT demonstrates promising screening performance as an initial single-center validation in RT patients. Multicenter confirmation and prospective evaluation against hard clinical endpoints are warranted.